[PATCH for v4l-utils] qv4l2: fix UVC support

[PATCH for v4l-utils] qv4l2: fix UVC support

[Hans Verkuil]

Hans,

Can you apply this to v4l-utils?

Thanks!

	Hans

>From 977eb253a3e7a5257a2051da240314e2dac385bf Mon Sep 17 00:00:00 2001
From: Hans Verkuil <hverkuil@xs4all.nl>
Date: Sat, 13 Mar 2010 23:19:48 +0100
Subject: [PATCH] qv4l2: fix UVC support

The workaround for drivers using videobuf broke the UVC support.
Fix this.

Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
---
 utils/qv4l2-qt4/general-tab.cpp |   18 +++++++++++++++---
 utils/qv4l2-qt4/qv4l2.cpp       |    4 +++-
 2 files changed, 18 insertions(+), 4 deletions(-)

diff --git a/utils/qv4l2-qt4/general-tab.cpp b/utils/qv4l2-qt4/general-tab.cpp
index 50a3edc..706b15d 100644
--- a/utils/qv4l2-qt4/general-tab.cpp
+++ b/utils/qv4l2-qt4/general-tab.cpp
@@ -198,11 +198,23 @@ GeneralTab::GeneralTab(const QString &device, v4l2 &fd, int n, QWidget *parent)
 	if (m_querycap.capabilities & V4L2_CAP_STREAMING) {
 		v4l2_requestbuffers reqbuf;
 
-		if (reqbufs_user_cap(reqbuf, 1))
+		// Yuck. The videobuf framework does not accept a count of 0.
+		// This is out-of-spec, but it means that the only way to test which
+		// method is supported is to give it a non-zero count. But non-videobuf
+		// drivers like uvc do not allow e.g. S_FMT calls after a REQBUFS call
+		// with non-zero counts unless there is a REQBUFS call with count == 0
+		// in between. This is actual proper behavior, although somewhat
+		// unexpected. So the only way at the moment to do this that works
+		// everywhere is to call REQBUFS with a count of 1, and then again with
+	       	// a count of 0.
+		if (reqbufs_user_cap(reqbuf, 1)) {
 			m_capMethods->addItem("User pointer I/O", QVariant(methodUser));
-
-		if (reqbufs_mmap_cap(reqbuf, 1))
+			reqbufs_user_cap(reqbuf, 0);
+		}
+		if (reqbufs_mmap_cap(reqbuf, 1)) {
 			m_capMethods->addItem("Memory mapped I/O", QVariant(methodMmap));
+			reqbufs_mmap_cap(reqbuf, 0);
+		}
 	}
 	if (m_querycap.capabilities & V4L2_CAP_READWRITE) {
 		m_capMethods->addItem("read()", QVariant(methodRead));
diff --git a/utils/qv4l2-qt4/qv4l2.cpp b/utils/qv4l2-qt4/qv4l2.cpp
index 96cb4d7..63881be 100644
--- a/utils/qv4l2-qt4/qv4l2.cpp
+++ b/utils/qv4l2-qt4/qv4l2.cpp
@@ -352,7 +352,7 @@ void ApplicationWindow::stopCapture()
 			if (-1 == munmap(m_buffers[i].start, m_buffers[i].length))
 				perror("munmap");
 		// Free all buffers.
-		reqbufs_mmap_out(reqbufs, 0);
+		reqbufs_mmap_cap(reqbufs, 0);
 		break;
 
 	case methodUser:
@@ -360,6 +360,8 @@ void ApplicationWindow::stopCapture()
 			perror("VIDIOC_STREAMOFF");
 		for (i = 0; i < m_nbuffers; ++i)
 			free(m_buffers[i].start);
+		// Free all buffers.
+		reqbufs_user_cap(reqbufs, 0);
 		break;
 	}
 	free(m_buffers);
-- 
1.6.4.2


-- 
Hans Verkuil - video4linux developer - sponsored by TANDBERG

Re: [PATCH for v4l-utils] qv4l2: fix UVC support

[Mauro Carvalho Chehab]

Hans Verkuil wrote:
> Hans,
> 
> Can you apply this to v4l-utils?

Hans V.,

You have access to v4l-utils. It would be nice if you could try to
apply it directly, for us to double check if the server is properly
working with a shared repository.

-- 

Cheers,
Mauro

Re: [PATCH for v4l-utils] qv4l2: fix UVC support

[Hans Verkuil]

On Sunday 14 March 2010 00:14:15 Mauro Carvalho Chehab wrote:
> Hans Verkuil wrote:
> > Hans,
> > 
> > Can you apply this to v4l-utils?
> 
> Hans V.,
> 
> You have access to v4l-utils. It would be nice if you could try to
> apply it directly, for us to double check if the server is properly
> working with a shared repository.

OK, done. It seems to work well.

	Hans

> 
> 

-- 
Hans Verkuil - video4linux developer - sponsored by TANDBERG